Field of the Disclosure
The present disclosure relates to the technical field of vision, and in particular, to a method and a device for acquiring depth information and a gesture recognition apparatus.
Description of the Related Art
As smart electronic devices increase and human-computer interaction technology develops, gesture interaction technology has gradually become a research focus of the people. Gesture interaction uses technology such as computer graphics to recognize body language of humans and convert it into commands to operate devices. Gesture interaction is a new human-computer interaction means other than a mouse, a keyboard and a touch screen. As virtual reality augmented reality (VRAR) technology becomes so popular, the gesture interaction technology will develop quickly in the coming years.
For gesture interaction technology, gesture recognition is a key technique. The related art has proposed a depth information-based gesture recognition method which needs to acquire the depth information of a target in field of view. At present, the depth information is acquired mainly by the following three methods, that is, a multi-camera vision method, a time of fly (TOF) method and a structure light method. In the multi-camera vision method, three-dimensional vision of two lenses is used to measure depth of scene. Cameras emit active light waves to illuminate three-dimensional scene and the light waves are reflected by the three-dimensional scene back to a depth camera. The accuracy of acquiring the depth information is reduced rapidly as the distance from the target increases and depends on states of surface of the target, in particular, when the target has less texture, the recognition accuracy will become relatively low. In the time of fly method, the time difference (phase difference) between time at which the light waves are emitted and the time at which the light waves are reflected is used to acquire the depth information of the three-dimensional scene. It has disadvantages of such as low resolution, inaccuracy of edge. The basic principle of the structure light method is to project one or more coded patterns into a target scene and the spatial coordinates of the object to be illuminated can be derived by means of calculating the deformation or relative change of displacement of information of coded structure light received by a camera device. The structure light method is a kind of efficient measurement method, but it needs to code and decode the projected patterns. Thus, it has a complex scheme in whole and high complicated algorithm and great calculation. In this way, the speed of generating the depth information is relatively slow such that the depth information cannot be acquired in real-time indeed. Meanwhile, the depth cameras which are based on these methods have complex optical and electronic devices and thus it is often difficult to be miniaturized and is expensive such that ordinary customers cannot accept them.